Power Grid Disturbance Record Grouping Using AI Waveform Features

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods, such as IEEE COMTRADE, lack the ability to automatically distinguish between real and test disturbance records (DRs) in power grids, necessitating a time-consuming and costly manual segregation process, especially when analyzing DRs from different regions/stations.

Innovation Solution

An AI-based approach using convolutional neural networks (CNN) or vision transformers to automatically segregate DRs into clusters based on waveform similarity and distortion, assigning labels to these clusters, and determining standard deviations for classification.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If manual segregation of disturbance records is performed, then accuracy in distinguishing real and test DRs is improved, but time consumption and cost increase significantly

Engineering Contradiction:
Improveaccuracy of DR segregationVSAvoidtime consumption for manual segregation
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent replaces the manual mechanical review process with an automated AI-based system. The convolutional neural network automatically analyzes waveform images of disturbance records, extracting features and making classification decisions without human intervention, thus eliminating time consumption while maintaining reliability through sophisticated pattern recognition

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system enables self-service classification where the AI model autonomously processes disturbance records, extracts relevant features from waveform images, and automatically assigns labels without requiring manual review. The system serves itself by having the CNN model perform both feature extraction and classification tasks independently

Inventive Principle:
Principle #25Self-service

2Adaptability or versatility

If manual review of disturbance records is performed at station level, then local knowledge utilization is improved, but central availability of segregated DRs deteriorates

Engineering Contradiction:
Improvelocal knowledge utilizationVSAvoidcentral availability of segregated DRs
Core Design Contradiction:
Adaptability or versatilityVSLoss of information

Solution Approach 1:

The AI-based system provides universal functionality that can process disturbance records from any station with consistent performance. The centralized AI server receives waveform images from multiple sources, processes them uniformly through the same CNN model, and returns standardized classifications, ensuring both local adaptability and central information availability simultaneously

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Productivity

If AI-based automated segregation is implemented, then productivity and efficiency are improved, but system complexity increases

Engineering Contradiction:
Improveefficiency of DR processingVSAvoidcomplexity of AI system
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The system segments the complex AI processing task into distinct functional modules: waveform image generation from disturbance records, feature extraction through convolutional layers, classification through fully connected layers, and result output. This modular architecture manages complexity by organizing the AI system into manageable, independent components that can be developed and maintained separately

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4618347A1Method, device and computer-readable storage medium for grouping disturbance records
Publication Date: 2025.09.17 HITACHI ENERGY LTD
  • EP4618347A1 patent drawingFigure 1A~1B
  • EP4618347A1 patent drawingFigure 2
  • EP4618347A1 patent drawingFigure 3

AI summary

The present disclosure relates to a method for grouping disturbance records, DRs, in a power grid. The method comprises obtaining a plurality of DRs, obtaining waveforms of the plurality of DRs, inputting the waveforms into an artificial intelligence, Al, model to extract a plurality of features, and grouping the DRs into at least a first group or a second group based on the extracted features.